Risk analysis,
probability and decision
tree analysis
PROF. DR. VEDAT CEYHAN
Risk and uncertainty
Risk
- Probabilities of possible outcomes are known
- Reflects the uncertain outcomes
- Negative occurance can be estimated,
Uncertainty
- Probabilities of possible outcomes are not known.
- Lacking of information
- Neither the probability nor the mode of
occurrence is known.
Risk and uncertainty
Uncertainty Risk and uncertainty Risk
Occurrence and Occurrence and Occurrence and
probability probability area partially probability
are not known known are known
Risk sources
Production risk
Market riski (price riski) Activities risk
Instituonal risk
Personal risk
Financial risk
Risk types
Pure risk
Measure of variaiton around the mean value
Downside risk
risk of a loss in particular, as opposed to
the symmetrical likelihood of a loss or gain
Upside risk
exposure to loss due to the pursuit of gain
can be thought of as “upside risk”..
Risk parties in agriculture
Farmers
Personal in agricultural institution
Input sales firm
Agricultural output sales firm
Policy maker and planners
Risk Management
Systematic activities for monitoring,
measuring, analyzing and defining risk
and developing strategies.
Best way to avoid from lost and to
benefit opportunities.
Steps of risk management
Establishing the context
Defining the risk
Analyzing the risk
Determining the risk strategies
Monitoring and evaluation
Establishing the context
Problem statement
Determining the risk related parameters
Situation may be strategic, institutional
or risk management
Defining risk
The list of occurrence affected the
performance
Determination of the answer of what,
how and why questions.
How farm affected from the
occurrence.
Focusing on measurable and
manageable risk.
Risk analysis (1)
Measuring probability of occurrence
Determination of the results of risk
management strategy
Risk analysis (2)
Quantifying risk
1. Data collection from different
sources
2. Measuring variation
Standard deviation
Variation coefficient
3. Determining probability and
probability distribution
Concept related risk analysis
Expected value (EV)
EV=probability x value
Certainty equivalent (CE)
the guaranteed amount of cash that a
person would consider as having the same
amount of desirability as a risky asset
Risk Premium (RP)
RP= EV– CE
Required calculation for risk quantifying
Probabilities
Probability distribution
Possible outcomes
Probability (as a measure of risk)
Bernoulli : measure of violence of
expectation on future prospect (objective
probability)
Keynes : probability should be defined by
using perception or feelings (subjective
probability)
What is the certain definition of risk?
Probability approaches
Classical probability approach
Relative frequency approach
Subjective probability approach
Classical probability approach
Based on symetry condition
Probability of occurrence are equal.
Case of dice: probability equals 1/6.
possible outcomes ofA
P(Ei)= ———————————————
total possible outcomes
Relative frequency approach
Based on counting occurrence
If trial is “n” times, A occurs “f” times;
P A
f
n
ATTENTION: Relative frequency is not probability,
it is approach.
Trial number must be sufficiently large.
Trial must be independent.
Subjective probability approach
Based on perception.
Reflects degree of individual beliefs.
“0” means disbelieve, “1” means absolute
beliefs.
Individual statements are accepted as a
probability.
Probability distribution
Presents the probabilities and forms the
distributions.
Show probabilities of all the possible
outcomes.
Have 4 basic characteristics such as mean,
variance, skewness and kurtosis.
Depict 3 different type such as table,
graphical presentation and mathematical
equation.
Random variable
Take the specific value with a spcific
probability.
- Yields
- Income
- Hail
- Loss sourced from dead
- Egg loss
Probability distributions
Discrete probability distribution
Continuous probability distribution
Discrete probability distribution
Characteristics
(i) vary between 0 and 1
(ii) the sum equals 1.
The mean value equals expected value.
E x xP x
Discrete probability distribution
Binomial distribution
Two possible outcomes: desirable and undesirable. Three
features of binomial trial are: (i) independent trial, (ii) two
possible outcomes and (iii) probability is fixed during the trial
Poisson distribution
n is too large, probability of occurrence is too small
Hypergeometric distribution
reflects the probability of occurrence in dependent trial.
Continuous probability distribution
All data in ratio level and continious.
Vary from 0 to 1
The sum of probability equals 1.
The probability of area between two point is
calculated.
The probability of one point is zero.
Continuous probability distribution
Normal distribution,
Normal approach to binomial dist.
Uniform distribution,
Exponential distribution
Eliciting probability distribution
Repeated occurrence (measurement is easy)
Classical probability approach
Relative frequency approach
Unrepeated occurrence (measurement is
difficult)
Subjective probability distribution
1. Reference gamble
2. Fractile method
3. Bayes analysis
Risk analysis method
Decision tree
Decision matrix
Mathematical models
Simulation techniques (Monte Carlo,
fuzzy logic etc.)
Decision tree
STEPS
Defining activities
Determining the level of activities,
Determining the critical factors and
calculating their probabilities,
Calculating the expected value for
each risk level
Decision tree example 1
Alternatives and their level
- Selling potatoes with current price
10000 TL (no risk)
- Storing potatoes and sell in future with
high price
Additional storage cost by 1000 TL
1. Potatoes supply is normal in future
(9500 – 1000 = 8500 TL)
2. Potatoes supply is low in future
(16000 – 1000 = 15000 TL)
Decision tree example 1
Selling now
10000 TL
Potatoes marketing
problem
Normal
supply
storage 8500 TL
Low
supply 15000 TL
Decision tree example 1
Selling now
10000 TL
Potatoes marketing problem
Normal supply
(0.75)
Storage 8500 TL
Low supply (0.25)
15000 TL
(0.75)(8500) + (0.25)(15000) = 10125 TL ,
Preference should be storage
Decision tree example 2
Cattle fattening in open area
Level of activities are 300, 400 or 500
cattle.
Climate good %20
normal %50
bad %30
OBD=12.2
expected value : 5.2 7 0
net revenue; 26 14 0
probability: 0.2 0.5 0.3
OBD = 10.8 OBD =11.3
4 5 1.8 expected value 6.8 7.5 -3
20 10 6 net revenue 34 15 - 10
0.2 0.5 0.3 ihtimaller 0.2 0.5 0.3
alternatieve investment : 300 400 500
Şekil Besi Sığırcılığı Örneğinin Karar Ağacı
Decision tree example 2
Decision criteria
- Mean highest expected value (400)
- Selecting highest revenue among alternative
having highest probabilities (500)
- Comparing risk and mean expected value
(400 / 500 = 400)
- Selecting highest revenue among alternative
having worst probabilities (300)
Payoff matrix
Use the same data with decision tree.
Data are presented in table format.
Event in first column, probability in
second column and net revenue in third
column.
Rows represents the expected value,
extreme values and interval.
Payoff matrix for example 1
Market condition Probability Expected value
Storage 10125
Normal supply 0.75 (8500 x 0.75) 6375
Low supply 0.25 (15000 x 0.25) 3750
Selling now 1.00 10000
Payoff matrix for example 2
Climate Probability Alternatives
condition 300 cattle 400 cattle 500 cattle
Good 0.2 20000 26000 34000
Normal 0.5 10000 14000 15000
Bad 0.3 6000 0 - 10000
Mean expected value (TL) 10800 12200 11300
Minimum value (TL) 6000 0 - 10000
Maximum value (TL) 20000 26000 34000
Interval (TL) 14000 26000 44000
Designing risk strategies
Defining objectives
Determining risk sources
Quantifying risk
Determining risk strategies
Risk strategies
Risk transfer
Risk taking
Risk control
Avoiding risk
Selecting strategies
Risk exposure Probability
low high
Small Risk taking Risk controlling
Large Risk transfer Avoiding risk
Risk control ve transfer measures (1)
Production
- Avoiding risky product
- Product diversification
- Insurance
- Renting
Marketing
- Contracts
- Cooperatives
- Forward agreement
- Minimum price agreement
- Hedging
- Price quarantee
Risk control ve transfer measures (2)
Finance
- Increasing liquidity
- Increasing solvency
Comparing risk strategies
Safety-first approach
ROY criteria and Telser criteria (MOTAD, TARGET
MOTAD etc.)
Expected variance approach
Linear risk programming, quadratic programming
Stochastic dominance approach)
Risk attitudes
Risk takers
certainty equivalent < expected value,
risk premium = +
Risk indifference
certainty equivalent = expected value
risk premium = 0
Riske averse
certainty equivalent > expected value
risk premium = -